《计算机应用》唯一官方网站 ›› 2026, Vol. 46 ›› Issue (8): 2447-2456.DOI: 10.11772/j.issn.1001-9081.2025070818

• 人工智能 • 上一篇    下一篇

相似度感知的民族文化知识图谱链路预测模型

王岩1,2, 陈曦1,2(), 赵中恺1,2, 周欢1,2, 吴涛3, 艾梦格1,2, 肖雪松4   

  1. 1.西南民族大学 计算机与人工智能学院,成都 610225
    2.计算机系统国家民委重点实验室(西南民族大学),成都 610225
    3.成都信息工程大学 计算机学院,成都 610225
    4.成都明途科技有限公司,成都 610213
  • 收稿日期:2025-07-22 修回日期:2025-09-30 接受日期:2025-10-14 发布日期:2025-11-05 出版日期:2026-08-10
  • 通讯作者: 陈曦
  • 作者简介:王岩(2000—),男,四川成都人,硕士研究生,主要研究方向:知识图谱、图神经网络
    陈曦(1985—),男,重庆人,教授,博士,主要研究方向:网络虚拟化、人工智能
    赵中恺(1993—),男,山西大同人,博士,主要研究方向:人工智能
    周欢(1990—),女,四川遂宁人,副教授,博士,主要研究方向:数据库、人工智能
    吴涛(1984—),女,山东曲阜人,教授,博士,主要研究方向:粒子群优化算法、人工智能
    艾梦格(2001—),女,河南周口人,硕士研究生,主要研究方向:图神经网络
    肖雪松(1975—),男,四川成都人,高级工程师,主要研究方向:智能体。
  • 基金资助:
    国家社会科学基金哲学社会科学领军人才项目(20230037);国家民委中青年英才项目(民委发〔2022〕117号);成都市科技项目(2025-XT00-00005-GX)

Similarity-aware link prediction model for ethnic culture knowledge graph

Yan WANG1,2, Xi CHEN1,2(), Zhongkai ZHAO1,2, Huan ZHOU1,2, Tao WU3, Mengge AI1,2, Xuesong XIAO4   

  1. 1.College of Computer Science and Artificial Intelligence,Southwest Minzu University,Chengdu Sichuan 610225,China
    2.The Key Laboratory for Computer Systems of State Ethnic Affairs Commission (Southwest Minzu University),Chengdu Sichuan 610225,China
    3.School of Computer Science,Chengdu University of Information Technology,Chengdu Sichuan 610225,China
    4.Chengdu Minto Technology Company Limited,Chengdu Sichuan 610213,China
  • Received:2025-07-22 Revised:2025-09-30 Accepted:2025-10-14 Online:2025-11-05 Published:2026-08-10
  • Contact: Xi CHEN
  • About author:WANG Yan, born in 2000, M. S. candidate. His research interests include knowledge graph, graph neural networks.
    ZHAO Zhongkai, born in 1993, Ph. D. His research interests include artificial intelligence.
    ZHOU Huan, born in 1990, Ph. D., associate professor. Her research interests include database, artificial intelligence.
    WU Tao, born in 1984, Ph. D., professor. Her research interests include particle swarm optimization algorithms, artificial intelligence.
    AI Mengge, born in 2001, M. S. candidate. Her research interests include graph neural networks.
    XIAO Xuesong, born in 1975, senior engineer. His research interests include agent.
  • Supported by:
    Leading Talent Project of National Social Science Fund of China(20230037);Science and Technology Project of Chengdu(2025-XT00-00005-GX);Young and Middle-aged Talent Project of National Ethnic Affairs Commission of China (〔2022〕117)

摘要:

挖掘民族文化知识图谱中实体间的潜在关联不仅有助于揭示民族文化要素之间的演化关系,也为民族文化的系统建模与智能推理提供了新的技术路径;然而,民族文化知识图谱存在多对一属性归属、结构高度同质化及存在大量噪声边等特点,现有的链路预测模型难以同时捕捉文化实体间的细粒度特征关联,并有效抵御图谱中的噪声边干扰,导致预测性能和模型鲁棒性受限。针对上述问题,提出一种特征相似度感知的图神经网络(GNN)模型SRGCN (Similarity-aware Relational Graph Convolutional Network)。SRGCN基于节点特征相似度构建动态聚合机制,以更准确地捕捉实体间的特征关联;同时,引入双层对比学习框架有效抑制噪声干扰,并设计线性加权的多目标损失函数,通过动态调整主、辅任务的损失权重,进一步增强模型的稳健性。在民族文化知识图谱数据集HeritEdge上的实验结果表明,在链路预测任务中相较于最优基线模型LTRGN (Linear self-aTtention with multi-Relational Graph Network),SRGCN在平均倒数排名(MRR)和Hits@10指标上分别提升28.4%和32.5%,性能更优。

关键词: 民族文化, 知识图谱, 图神经网络, 链路预测, 对比学习

Abstract:

Mining potential associations among entities in ethnic culture knowledge graphs is valuable for revealing evolutionary relationships among cultural elements and provides new technical paths for systematic modeling and intelligent reasoning of ethnic culture. However, because such graphs have characteristics of many-to-one attribute affiliations, highly homogeneous structures, and numerous noisy edges, the existing link prediction models are limited in capturing fine-grained feature associations among cultural entities simultaneously and resisting noisy edge interference in the graphs effectively, which restricts prediction performance and model robustness. To address these problems, a Graph Neural Network (GNN) model with similarity-aware features, SRGCN (Similarity-aware Relational Graph Convolutional Network) was proposed. In SRGCN, a dynamic aggregation mechanism was constructed on the basis of node feature similarity, so as to capture feature associations among entities more accurately. At the same time, a dual-level contrastive learning framework was introduced to suppress noise interference effectively, and a linearly weighted multi-objective loss function was designed, where the weights of primary and auxiliary tasks were adjusted dynamically to further enhance the model robustness. Experimental results on the HeritEdge ethnic culture knowledge graph dataset show that SRGCN outperforms the optimal baseline model, LTRGN (Linear self-attention with multi-Relational Graph Network), achieving improvements of 28.4% and 32.5% on Mean Reciprocal Rank (MRR) and Hits@10, respectively, demonstrating better performance.

Key words: ethnic culture, knowledge graph, Graph Neural Network (GNN), link prediction, contrastive learning

中图分类号: